Abstract: As product reviews in the internet are helpful for the decision of online shopping,and the classification accuracy of sentiment analysis is one of important problems．Recently,ensemble learning has been proved to be an effective method of enhancing the classification accuracy．Bagging and Boosting have been applied into the sentiment analysis,while Random Subspace is paid less attention to．In this paper,an new method,RS-SVM,was proposed for sentiment analysis based on the characteristic of high dimension of product review''s dataset．RS-SVM uses the state-of-the-art SVM as base learner and Random Subspace as ensemble method in order to enhance the accuracy of sentiment analysis．Lastly,experiments based on movie reviews'' dataset were conducted to verify the effectiveness of RS-SVM．Experimental results reveal that RS-SVM gets the best classification results compared with other methods.

Key words: Sentiment analysis,Production review,Ensemble learning,Random subspace,SVM